MLReal: Bridging the gap between training on synthetic data and real data applications in machine learning
This repository accompanies publication [Alkhalifah, Wang and Ovcharenko, 2021] and provides forward / backword methods for MLReal transformation in application to the task of low-frequency data extrapolation.
Download the data and run notebooks. All notebooks are set for inference / view by default. Meaning that these will not run any heavy calculations unless reset otherwise. Instead, these will use the pre-trained weights and data to partially reproduce results from the paper.
| Filename | Description |
|---|---|
| ex0_create_training_dataset.ipynb | Generate training dataset of synthetic waveforms |
| ex1_multi.ipynb | Train Multi-column network to predict low-frequency data |
| ex2_mlreal_forw.ipynb | Same as ex1 but with MLReal applied to inputs only |
| ex3_mlreal_forw_back.ipynb | Same as ex1 but with MLReal applied to both inputs and targets |
| shared_data_loading.ipynb | Shared data loading snippets |
| assets | Folder with images for README |
| pretrained_files | Download and place pre-trained data here |
| utils | Code components |
- Python 3.8
- PyTorch 1.8
- CUDA 11.0
For the rest of Python dependencies check requirements.txt.
git clone https://github.com/swag-kaust/mlreal.git
cd mlreal/
python -m venv env
source env/bin/activate
pip install -r requirements.txt
jupyter notebook .
Unzip files by running tar -xvf arhive.tar.gz and place complete folders according to the table
| Link | Size | Destination | Description |
|---|---|---|---|
| data.tar.gz | ~ 13 Gb | ./pretrained_files/data/* |
training and validation datasets |
| trained_nets.tar.gz | ~ 300 Mb | ./pretrained_files/trained_nets/* |
Pre-trained network weights |
Our implementation is heavily influenced and contains code blocks from Inpainting GMCNN.
@misc{alkhalifah2021mlreal,
title={MLReal: Bridging the gap between training on synthetic data and real data applications in machine learning},
author={Tariq Alkhalifah and Hanchen Wang and Oleg Ovcharenko},
year={2021},
eprint={2109.05294},
archivePrefix={arXiv},
primaryClass={physics.geo-ph}